PBL Height Retrievals at a Coastal Site Using Multi-Instrument Profiling Methods

Abstract

The objective of this study was the estimation of the dynamic evolution of the Planetary Boundary Layer (PBL) height, using advanced remote sensing measurements from Finokalia Station, where the Pre-TECT Campaign took place during 1–26 April 2017. PollyXT Raman Lidar and Halo Wind Doppler Lidar profiles were used to study the daily vertical evolution of the PBL. Wavelet Covariance Transform (WCT) and Threshold Method (TM) were performed on different products acquired from Lidars. According to the analysis, all methods and products are able to provide reasonable boundary-layer height estimates, each of them showing assets and barriers under certain conditions. Two cases are presented in detail, indicating the limited daytime evolution of a coastal area, the decisive role of wind speed-direction in the formation of a shallow or high boundary layer and the differences when using aerosols or turbulence as tracers for the PBL height retrieval. Comparison between the observed PBL and ECMWF model results was made, establishing the importance of actual PBL measurements, in coastal regions with complex topography.

1. Introduction

The Planetary Boundary Layer (PBL) height estimation provides useful information on the available air volume above the Earth’s surface into which emitted air pollutants are diluted and thus is a key parameter in air quality and weather or dispersion forecasting models. Over land surfaces under high-pressure conditions, the boundary layer has a well-defined structure that evolves with the diurnal cycle [1]. The three major components of this structure are the Mixed Layer (ML), the residual layer, and the stable boundary layer. When it comes to coastal regions the daytime evolution can be less pronounced and the formation of an Internal Boundary Layer (IBL) is a common phenomenon. During a sunny day, the sea breeze will blow the stable or neutrally stratified air over the sea towards land. The surface heating effect and dynamic disturbance effect intensify the turbulence in the lowest atmospheric layer to form an unstable layer, which develops into the Thermal Internal Boundary Layer (TIBL) [2]. Also, an IBL can develop when flow over a uniform surface encounters a step-change in surface conditions such as aerodynamic roughness, temperature, or moisture [3]. Moreover, the TIBL characteristics over the coastal land after the onset of the sea breeze are similar to those of the shallow convective boundary layer, commonly observed over plain areas. Therefore, the height of the TIBL can be considered the height of the shallow convective boundary layer [2]. In the case of an offshore wind flow, the internal structure of the IBL can present many similarities with the slowly evolving stable boundary layer overland [4].

Studying PBL in regions with special topography such as steep cliffs, islands, or mountains, highlights the importance of field measurements. In this analysis, the PBL diurnal evolution and characteristics are examined at Finokalia, a coastal region on the northern coast of Crete, Greece, located at the top of a hilly elevation, facing the sea within a sector 270° to 90°. The location of the island, along with the atmospheric circulation patterns [5], form a rich aerosol scene in the area, originating from natural sources (marine aerosols from the Aegean Sea, dust particles from Africa) or anthropogenic activities (pollution from big regional centers such as Istanbul and Heraklion). On a synoptic scale, the Eastern Mediterranean is at the crossroads of air mass outflows from European and Asian pollution sources [6] and receives significant amounts of desert dust from Africa and the Middle East. The region is characterized by considerable variability in cloud systems, ranging from frontal and convective to cyclones [7,8,9]. It is a climate “hot spot”, exhibiting more frequent and more intense weather phenomena associated with severe winds (Etesians), floods, and dust events during the transition seasons [10,11,12,13,14]. Therefore, accurate PBL height is needed to correctly interpret the numerous measurements at the Finokalia site.

The use of Lidar systems for active remote sensing of meteorological parameters and aerosols, is an effective way to detect ML height. Past studies have verified PBL height obtained from Lidar data by using several methods, against PBL height derived from radiosonde data [15,16,17,18,19,20]. Studying ML by means of Lidars, can suffer from many restrictions related to weather conditions, range, and accuracy, but can also provide high temporal and spatial resolution. The retrieval of the PBL top in a region with special topography, can hardly be an automated procedure, because of the variability inserted by aerosols and cloud presence, as well as by strong wind fields. Furthermore, very shallow PBL heights can occur at coastal locations [21]. Baars et al. [15] describe the continuous monitoring of the boundary-layer top with lidars, using several modifications of the Wavelet Covariance Transform, according to the different backscatter conditions in rather clean or very polluted air.

In this study, measurements from collocated PollyXT Raman Lidar and Halo Wind Doppler Lidar, collected during the PreTECT Campaign, were used to investigate the PBL dynamic characteristics. The PreTECT experimental Campaign (described in Section 2) took place in Finokalia, Crete, where remote sensing and in-situ measurements were obtained. Wavelet Covariance Transform (WCT) and Threshold Method (TM), are applied on Lidar products, that work as indicators for the ML. This work contributes to the discussion of PBL evolution and height detection [22,23], comparing methods applied to different metrics for two case studies characterized by different meteorological conditions and through month-long analysis. Comparing field measurements with model results, contributes to the fundamental understanding of the PBL [24,25,26]. In our study, the observed PBL form Lidars is compared with the ERA5 Re-Analysis dataset from European Centre for Medium-Range Weather Forecasts (ECMWF) results.

The manuscript is structured as follows: Section 2 describes the instrumentation, data, and the methodology applied to retrieve the PBL height. Section 3 presents the case studies and the associated meteorological conditions analyzed in detail. The results are discussed along with the statistical analysis performed for PBL height retrievals during the campaign. The main conclusions are summarized in Section 4.

2. Instrumentation, Data, and Methodology

In this study, an ensemble of active remote sensing instrument observations is used in order to derive the PBL height. As mentioned in the introduction, the measurements were collected in the frame of the Pre-TECT Campaign [27] that took place in Finokalia (35.34°N, 25.67°E, 250 m a.s.l.) (Figure 1), Crete, during the period 1–26 April 2017, organized by the National Observatory of Athens (NOA). Observations from PollyXT Raman Lidar [28] and a Halo scanning Wind Doppler Lidar [29], are also processed. Wavelet Covariance Transform and Threshold Method are applied with several modifications to the products of the campaign dataset.

2.1. Instrumentation

This section gives an overview of the two active remote sensing instruments operating during the PreTECT campaign.

The multi-wavelength depolarization, Raman PollyXT Lidar of the National Observatory of Athens (NOA), was built in 2014 and operated in Athens (2015–2016), Nicosia (for 2 campaigns in March 2015 and April 2016), Finokalia (2017–2018) and Antikythera (2018–today). The system components are briefly depicted in

Table 1. Specifications of PollyXT Raman Lidar, more information at [28].

PollyXT Lidar Specifications
Operating Wavelengths	355 nm, 532 nm, 1064 nm
Number of channels	12
Height resolution	7.5 m
Time resolution	30 s
Maximum Altitude	40 km
Pre-Trigger	256 bins
Repetition rate of laser	20 Hz
Photomultiplier voltage	127 Volts
Zenith angle of measurement	5°
Polarization	Cross and Total

, but a more detailed description can be found in [28]. PollyXT Lidar consists of a compact, pulsed Nd:YAG laser, emitting at 355, 532, and 1064 nm at a 20 Hz repetition rate, with the laser beam pointed into the atmosphere at an off-zenith angle of 5°. The backscattered signal is collected by a Newtonian telescope with a 0.9 m focal length, acquiring profiles with a vertical resolution of 7.5 m, and a temporal resolution of 30 s. The system includes five elastic channels for far and near range detection at 355, 532, and 1064 nm; four Raman channels for far and near range detection at 387 and 607 nm; two depolarization channels at 355 nm and 532 nm; and a water vapor channel at 407 nm. The overlap-affected height range of the overall system is 120 m above the lidar for the Near Field channels and between 700 and 800 m height for the Far Field channels. All measurements within the period 2014–2022 are available online at http://polly.tropos.de (accessed on 15 August 2022).

The Halo Photonics Stream Line Scanning Doppler lidar, is a 1.5 µm pulsed Doppler lidar with a heterodyne detector that can switch between co- and cross-polar channels [25]. The Doppler lidar has a range resolution of 30 m and measures the attenuated aerosol backscatter and Doppler velocity along the beam direction. Two conical scans at 15° and 75° elevation angle and a 3° elevation angle sector scan were scheduled every 15 min. This schedule leaves approx. 9 min out of every 15 min for vertically-pointing measurement. For the vertically-pointing measurements integration time was set to 3.7 s. For the 15° conical scan integration time was 4.5 s and for the 75° conical scan integration time was 6.5 s. More details about Halo Doppler Lidar characteristics and specifications can be found at [29,30]. The technical specifications of the instrument as configured for standard operation during the campaign are summarized in

Table 2. Specifications of Halo Wind Lidar, more information at [29].

Halo Wind Lidar Specifications
Wavelength	1.5 μm
Detector	heterodyne
Pulse repetition frequency	15 kHz
Nyquist velocity	20 ms−1
Sampling frequency	50 MHz
Velocity resolution	0.038 ms−1
Height resolution	30 m
Range	90–9600 m
Pulse duration	0.2 µs

.

2.2. Data

For the case studies analyzed herein, several products are retrieved from each instrument. Range Corrected Signal (RCS), attenuated BackSCatter coefficient (BSC) and Water Vapor Mixing Ratio (WVMR) are calculated from the PollyXT Lidar, using the signal from 532 nm Near Field (NF) and 1064 nm Far Field (FF) channel for the BSC product and the ratio of 387 nm and 407 nm signals for the WVMR product [31]. WVMR profiles from PollyXT lidar are calibrated by the collocated Microwave Radiometer (MWR) [32,33]. The WVMR retrievals are only available during nighttime hours, 00:00–04:00 and 18:00–24:00 UTC.

The Halo Doppler Lidar measurements were post-processed according to [30,34,35] and an SNR-threshold of 0.0075 was applied to the data. Horizontal wind speed and direction were retrieved from the conical scans [36] and turbulent kinetic energy dissipation rate (TKE_dr) was estimated following [37]. Both Lidars are placed at 252 m above sea level.

For the meteorological analysis of the case studies, the ERA5 Re-Analysis dataset from European Centre for Medium-Range Weather Forecasts (ECMWF) is used [38], at 0.25° × 0.25° resolution with 137 levels [39]. The ERA5 reanalysis model defines the PBL height as the minimum height for the bulk Richardson number reaching the critical value of 0.25, according to the algorithm that was proposed by Vogelezang and Holtslag [40] and is publicly available through the Copernicus Climate Data Store (CDS) [41]. For the vertical meteorological parameters needed for the retrieval of WVMR from PollyXT Lidar, we utilize the Advanced Research Weather Research and Forecasting model version 4.2.1 (WRF-ARW) [42]. The WRF-ARW spatial set up was at 9 × 9 km resolution domain with 600 × 370 grid points and 33 vertical levels. Simulations were initiated at 00:00 UTC on 1 April 2017 and were completed at 18:00 UTC on 30 April 2017. Finally, the atmospheric transport modeling used in this study, is based on the FLEXPART v10.4 (FLEXible PARTicle) Lagrangian dispersion model [43,44]. The following are reproduced:

(a)

8-day air masses back-trajectories prior to their arrival at 2 altitudes (0.5 and 5 km) above Finokalia ground station on 10 April 2017 at 10:00 UTC and

(b)

5-day air masses back-trajectories at 500 m above Finokalia ground station on 14 April 2017 at 10:00 UTC.

The footprint output from FLEXPART, is calculated in unit second (s) according to [44,45]. These specific periods, were chosen to describe the aerosol scene of the cases examined in detail (described in Section 3). The altitudes (500 m, 5 km), correspond to the observed aerosol layers, in order to provide information about their origin and trajectory.

A total of 50,000 particles are released over the Finokalia station in both simulations described above. FLEXPART was driven with 3-hourly meteorological data from the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) analyses provided at 0.5° × 0.5° resolution and for 41 model pressure levels.

2.3. Methodology

Wavelet Covariance Transform (WCT) is applied on 532 nm NF RCS and WVMR profiles acquired from PollyXT Lidar and on BSC profiles from Halo Lidar. RCS and BSC are proportional to the air concentration (molecules and aerosols). Threshold method (TM) is applied on TKE_dr profiles from Halo Lidar. All the product profiles used in this study, are time-averaged with a 15-min mean, with the exception of TKE_dr profiles where a 1-h median is calculated.

2.3.1. Wavelet Covariance Transform (WCT)

The WCT method applied to the above-mentioned products, assumes that the majority of aerosol particles are contained in the PBL, rather than in the free troposphere. Hence, a strong decrease of the signal is observable at ML top. The wavelet covariance function is calculated for every single averaged profile according to the following formula [46]:

$${\mathrm{W}}_{\mathrm{f}}\left(\mathrm{a},\mathrm{b}\right)=\frac{1}{\mathrm{a}}{\int }_{{\mathrm{z}}_{\mathrm{b}}}^{{\mathrm{z}}_{\mathrm{t}}}\mathrm{f}\left(\mathrm{z}\right)\mathrm{h}\left(\frac{\mathrm{z}-\mathrm{b}}{\mathrm{a}}\right)\mathrm{dz},$$

$$\mathrm{with}\mathrm{the}\mathrm{Haar}\mathrm{function}:\mathrm{h}\left(\frac{\mathrm{z}-\mathrm{b}}{\mathrm{a}}\right)=\{\begin{array}{c}+1, b-\frac{\mathrm{a}}{2}\le z\le b\\ -1, b\le z\le b+\frac{\mathrm{a}}{2}\\ 0 \mathrm{elsewhere}\end{array}$$

f(z) is the signal to perform WCT (in this case it is RCS, BSC, or calibrated WVMR), z is the measurement height, z_b and z_t are the lower and upper limits of the signal profile, respectively. The dilation a is the extent of the step function (window) and the translation b determines the location of the step.

A critical detail for the accurate WCT application, is the selection of an appropriate value of the dilation so as to distinguish PBL, cloud layers, and aerosol layers [15]. For the examined cases, sensitivity studies were performed, resulting in different dilations, and an algorithm was developed to detect the maxima of the time-averaged vertical profiles for each product. In Figure 2, an example of wavelet signal profiles of 14 April 2017 02:00 UTC is presented (blue color), along with the corresponding product from PollyXT and Halo Lidar, showing the detection of the peak around 400 m. The selection of a fixed dilation of 20·Δz = 150 m, where Δz corresponds to the lidar vertical resolution 7.5 m, works well for the cases of this study.

2.3.2. Threshold Method

The Threshold Method (TM) is based on a set of a condition that includes a constant threshold value. More specifically, the TM is applied on the 1-h median TKE_dr (ε) profiles as follows: starting from the lowest usable range gate, as high up as TKE_dr is larger than the threshold value, the range is considered to lie within the mixing layer. Therefore, the altitude corresponding to this threshold value is interpreted as the turbulent Mixing Layer Height (MLH). For our cases, MLH is determined using the threshold of 10⁻⁴ m²·s⁻³, as suggested in previous studies [21].

The products from each instrument and the method that is applied in each case, are presented in

Table 3. Instruments-Products-Methods, abbreviations are presented at the end of the manuscript.

Instrument	Product	Method
PollyXT Lidar	RCS	WCT
PollyXT Lidar	WVMR	WCT
Halo Wind Lidar	BSC	WCT
Halo Wind Lidar	TKEdr	TM

.

2.3.3. Comparison between WCT and TM Applied Methods

The methods used in this study in order to retrieve PBL height, present similarities, and differences. WCT detects a strong decrease of the signal of a parameter, while TM detects the first value that overpasses the threshold set by the user. The systematic differences that may arise during inter-comparison of the methods in this study, are attributed to the different parameters on which each method is applied. WCT is used for the detection of PBL height, using aerosols (RCS, BSC, WVMR) as tracers, while TM is used for the detection of PBLH, using a thermodynamic parameter (TKE_dr) as tracer. In Figure 3, profiles of products are presented, along with the method used, for 10 April 2017, 01:30 UTC. As observed, WCT captures a layer around 950 m, but TM detects the first point that is lower than the threshold at 790 m approximately. This difference is attributed to the scientific parameter that is used as an indicator for the PBL or MLH height.

3. Results

Before analyzing statistics of PBL height at Finokalia during the PreTECT Campaign, two days are presented with different meteorological conditions and aerosol mixtures. Firstly, 10 April 2017 is described by the presence of clouds and aerosols and secondly, 14 April 2017, is a day without clouds and significant aerosol load. In Figure 4, wind conditions are illustrated. For the first case, mostly N/NW moderate winds are dominant (Figure 4a) in addition to the second case, where W winds are present up to 600 m, turning to NW above (Figure 4b). The main wind flow in Finokalia is from the North, considering that it is an island in the south Aegean Sea, affected by the Etesians. The presence of Crete Island functions as an obstacle and shifts the wind to NW and W near the surface.

3.1. Case 10 April 2017

3.1.1. Meteorological Analysis

The existence of Azores high over the north Atlantic Ocean and a deep trough over the Black Sea forces the atmospheric mesoscale circulation in Eastern Balkans and therefore, the weather conditions in Greece on this day. The movement of the polar jet stream southwards in Europe, which started in 7 April 2017, permitted colder air masses to enter Eastern Europe and spin with the support of the associated trough system. As shown in Figure 4a, the low-pressure system is located over the Black sea at 500 hPa, accompanied by the colder air mass, presented at 850 hPa with a temperature of 0 °C (Figure 5b).

As a result, the weather in Crete, Greece is characterized by low temperatures, favoring cloud formation. The wind field in the Aegean Sea is mostly N/NE, stronger in the central and southern parts reaching up to 12 m/s (6 Beaufort-fresh to a strong breeze, Figure 5c).

On 10 April 2017, sparse high-level (8–10 km) and mid-level (5–8 km) clouds are observed above Finokalia Station, as shown in Figure 5a, depicting the attenuated BSC coefficient at 1064 nm (upper panel) and the volume depolarization ratio (bottom panel) from PollyXT Lidar. Low level (0–2 km) clouds also form inside and above the PBL. The thin aerosol layer observed at 5–6 km presents a depolarization ratio of 15–30% (Figure 6b) and according to the FLEXPART model, originates from NW Africa (green line Figure 6c,d) which suggests a mixture of dust and pollution aerosols. Moreover, marine and pollution aerosols with a volume depolarization ratio of 5–10% exist up to 700 m, coming from the Balkans and the Aegean Sea (blue line Figure 6c,d).

3.1.2. PBL Diurnal Evolution on 10 April 2017

Winds from the northern sector (NNW and NNE) are dominant during the entire day, as shown in Figure 4, stronger from 00:00 to 14:00 UTC (up to 8 m/s), compared to the afternoon. A shift in wind direction is observed after 18:00 UTC: west-northwesterly (W/NW) winds reach up to 300 m, switch to north-northeasterly (N/NE) above 700 m, similar to a land breeze.

PBL height is retrieved by applying WCT on the RCS from the 532 nm Near Field channel (Figure 7a—brown circles) and the WVMR (Figure 7c—green triangles) from PollyXT Lidar, as well as on the attenuated BSC from Halo Wind Lidar (Figure 7b—blue circles), resulting to PBL_RCS, PBL_WVMR, PBL_BSC, respectively. The TM is applied on the TKE_dr, provided by Halo (Figure 7d—red stars), resulting in MLH_TKE retrievals. At the beginning of the day (00:00–04:00 UTC), PBL_BSC, PBL_WVMR, and PBL_BSC, reach up to 1km and then drop gradually to 400 m until 03:00 UTC, whereas MLH_TKE, reaches 800 m and then drops to 300 m. During 03:00–11:00 UTC, PBL (where available) presents a rising tendency, from 400 m to 1 km, corresponding to the daytime evolution. After 11:00 UTC, the PBL_RCS and PBL_BSC is descending from 1 km to 600 m until 18:00 UTC, presenting some local peaks in the meantime. MLH_TKE follows that descending from 1 km to 500 m between 10:00–17:00 UTC. After 18:00 UTC, an aerosol layer with top ranging from 500 to 800 m is observed from WCT on RCS, WVMR, and BSC products.

In Figure 8, the PBL height and MLH retrievals for 10 April 2017 are compared. At the beginning (00:00–02:00 UTC), all approaches capture a high nocturnal mixing layer. This is probably the result of mechanical turbulence induced by the northerly winds, meeting the coastline. Aerosols and water vapor are transported above the site (as explained in Section 3.1.1), by the same flow and track the mixing layer indicated by TKE_dr. After 02:00 UTC, the descending tendency of PBL is captured from all products, as well as the daytime evolution that takes place after 04:00 UTC, when the sun rises and thermal turbulence starts to form. During daytime (05:00–16:00 UTC), PBL_RCS and PBL_BSC, present accordance in capturing a well-mixed layer reaching 1000 m, which is probably a result of both mechanical and thermal turbulence. However, comparing PBL_RCS, PBL_BSC, and PBL_WVMR with MLH_TKE result in some significant differences. PBL from all products, track a higher layer during the entire day, in addition to MLH_TKE which tracks the dynamically or thermally turbulent layer, which is significantly dropping after 18:00 UTC. According to [3], the region of significant turbulence is associated with a mesoscale circulation superimposed on the mean offshore flow, and probably relates to a sea-breeze-type circulation. In this case, MLH_TKE captures a TIBL that is developed from 02:00 to 17:00 UTC and presents many similarities to a convective boundary layer. Hence the local peak observed at 10:30 UTC is common for both PBL and MLH retrievals. The remarkable difference that is observed after 18:00 UTC, is attributed to the wind direction change, from west (W) winds at the lower layer (at 100 m a.g.l. approximately, Figure 4a) to the north (N) winds at the upper layer, indicating highly stable conditions. The layer near the surface from 18:00 UTC onwards, is directly affected by the land and becomes cooler during the night, while above 300 m the wind flow is from the sea (warmer) and contributes to the continuous transfer of aerosol load and water vapor above Finokalia, used as tracers from the WCT method. This is probably related to land-breeze-type flow, favoring the development of a stable IBL, represented by MLH_TKE. Also, WCT on aerosol tracers detects a layer that is rich in aerosols and probably coincides with the residual layer, without following the above-mentioned MLH_TKE drop (similar to e.g., [47]) that conceives a stable IBL. PBL from the other products reflects a layer, that is not affected by turbulent transport of surface-related properties and hence does not fall within the nocturnal boundary layer.

The PBL height estimated by the ECMWF model, is presented in Figure 8 for two points that belong in different grids: the purple diamond (PBL_ECMWFSEA) belongs to a grid simulated over the sea and the orange diamond (PBL_ECMWFLAND) belongs to a grid simulated over land. Finokalia Station, is located between the two model bins. It is noted that the PBL height derived from ERA5 ECMWF (see Section 2.2), is referred to as ground level, similarly to the PBL height derived from Lidars. During all day, there is good agreement between PBL_ECMWFSEA and PBL height measured by the instruments. PBL_ECMWFSEA does not present a sharp daytime evolution, in addition to PBL_ECMWFLAND which ascents between 04:00–10:00 UTC and descents after 14:00 UTC. Nighttime (00:00–04:00 UTC, 18:00–24:00 UTC) measurements of PBL height are in sufficient agreement with the PBL_ECMWFSEA. On the other hand, daytime measurements are found to agree mostly with the rising tendency that PBL_ECMWFLAND presents (04:00–08:00 UTC), but there is a discordance after 16:00 UTC, where PBL_ECMWFLAND drops very smoothly. MLH_TKE, also captures this drop, since the stable IBL is formed in common behavior as a stable nocturnal layer.

3.2. Case 14 April 2017

3.2.1. Meteorological Analysis

On 14 April 2017, zonal flow is restored above Europe, with the polar jet stream limited in the northern paths (500 hPa Figure 9a). Warmer air masses are approaching the Mediterranean Sea with temperatures ranging from 10–15 °C at the 850 hPa level (Figure 9b). Under these atmospheric conditions, the weather in Crete is sunny and warmer. The W/NW wind field that dominates the southern parts of the Aegean (Figure 9c), does not exceed 12 m/s all day.

Regarding the cloud and aerosol conditions on 14 April 2017, skies are cloudless all day above Finokalia station, according to the attenuated BSC coefficient derived from PollyXT Lidar (Figure 10a). The aerosol concentration is very low; probably a mixture of marine and pollution particles with a depolarization ratio of less than 10% (Figure 10b), originating from NW (Figure 10c–e), are present in the lower levels.

3.2.2. PBL Diurnal Evolution on 14 April 2017

The second case examined, 14 April 2017, is a day with much stronger winds, especially at the beginning of the day, in comparison with 10 April 2017. At the lower levels (below 500 m), the W wind field prevails, with a speed exceeding 15 m/s, in addition to the higher levels (above 500 m), affected by N/NW winds (Figure 4b). After 08:00 UTC, winds above 500 m decrease significantly, but the vertical difference in wind direction is maintained throughout the day. A low-level jet is present throughout the day with peak height at approx. 200 m a.g.l. Overall, the vertical profile of horizontal wind indicates persistent layering, which suggests different characteristics of the air masses at the surface and above 500 m a. g.l.

Before sunrise (00:00–04:00 UTC), WCT on RCS, BSC, and WVMR products, locate a layer around 400 m (Figure 11a–c), while MLH_TKE is captured between 200 and 350 m (Figure 11d). After 04:00 UTC, daytime evolution takes place and the PBL starts rising as captured by RCS and BSC products and then drops again after 06:00 UTC to 200 m with small fluctuations. TKE_dr from Halo, records a very turbulent layer below 400 m during all day, likely caused by mechanical turbulence generated by the W winds (Figure 4b). During the daytime (05:00 to 14:00 UTC), there is also another turbulent layer present, reaching from 400 m to up to 800 m a.g.l., but this layer seems to be unconnected to the surface, because of the decrease of turbulence at 400 m. Moreover, the lower detected layer is rich in water vapor, reaching around 400 m before sunrise and 200 m after sunset (Figure 11c). As described in Section 2.2, only nighttime measurements of water vapor are available.

Examining all the results together (Figure 12), indicates a significant accordance between all methods and retrievals with a slightly lower MLH_TKE. The vertical profile of each product presents a strong decrease, detected by WCT and TM, resulting in strong agreement between the different products. When the sun rises (after 04:00 UTC), weak daytime evolution takes place, with the PBL height and MLH_TKE captured by the two Lidars, being in sufficient agreement. Some profiles of Halo BSC and PollyXT RCS, capture a layer around 400 m after 16:00 UTC. This could happen because the northern winds that are imposed from the synoptic system, are modified by the surface due to the presence (obstacle) of Crete and slightly switch to western direction (Figure 4b). As a result, lifting marine and pollution aerosols on the site, are detected by WCT method and causing the outliers of PBL_RCS, PBL_BSC. Also the PBL_RCS and PBL_BSC anomalies after 06:00 UTC are very likely to be connected with the wind speed local minimum at 400 m, shown in Figure 3b. After 18:00 UTC, the MLH_TKE is mainly detected below 120 m, but PBL from all products is located around 200 m. This indicates the possibility of a stable TIBL formed when there is a slight weakening of wind speed below 200 m (Figure 4b).

Furthermore, comparing our results with the ECMWF retrievals, there is a slight underestimation of PBL_ECMWFLAND at the beginning of the day (00:00–04:00 UTC) and good agreement at the end (18:00 UTC onwards). The underestimation occurs because lidars detect aerosols trapped in the boundary layer, whereas the model predicts a lower nocturnal layer above land. On the other hand, MLH_TKE is closer to the PBL_ECMWFLAND at the beginning of the day, as it represents a lower turbulent layer. PBL_ECMWFSEA presents the same tendency as PBL_WVMR, PBL_RCS and PBL_BSC during 00:00–04:00 UTC, but deviates after 16:00 UTC, maintaining the values of the nocturnal layer (around 400 m). This overestimation of PBL_ECMWFSEA is closer to the outliers of PBL_BSC and PBL_RCS. Nevertheless, a substantial difference arises from 06:00 to 16:00 UTC, with the model’s overestimated PBL_ECMWFLAND related to lidar retrievals during the convective daytime period. As expected, PBL_ECMWFSEA is not developing as quickly and is in better accordance with the measurements, still overestimating the measured PBL during 07:00–16:00 UTC. This divergence between PBL_ECMWFSEA, PBL_ECMWFLAND and the measurements, especially during daytime, is attributed at the complexes induced by Finokalia’s location. The station is at the edge of a steep cliff between the land and sea, but the model conceives a land surface with high convective activity in case of PBL_ECMWFLAND and a maritime area in which the marine atmospheric boundary layer forms in case of PBL_ECMWFSEA. The influence of changes in surface roughness at Finokalia station, occurs at a scale that is too fine for the model, hence the above-mentioned differences arise. It is likely that the continuous on-shore flow yields a vertical profile over the coast on land that closely resembles the maritime near-shore environment, whereas the model’s surface fluxes are likely insensitive to this advection as the surface fluxes are typically one-dimensional based on surface properties.

3.3. Statistical Analysis

An overview of wind speed and direction during 1–26 April at Finokalia, is presented in Figure 13b,c, indicating variable meteorological conditions and aerosol transportation. In Figure 13a, the attenuated backscatter coefficient at 532 nm is displayed, along with the PBL_RCS. White parts of the plot, signify the presence of low clouds, and black parts stand for the attenuated signal above clouds. Time periods with low-level clouds were excluded from PBL retrievals. No sharp daytime evolution of PBL is observed and its height varies between 200 and 1000 m approximately. It is also shown that when N or NW winds are present over a wide portion of the vertical profile (1–3, 9–12 April), PBL is higher. W winds in the lower levels, favor a shallower PBL, (6–8, 13–16 April) and winds from the eastern and southern sectors, favor the cloud formation (3–6, 12, 21–23 April).

Almost 30% of wind measurements of the campaign on the lower levels (105 m) come from W, as shown in the wind rose in Figure 14a. At 525 m, the dominating direction slightly shifts to NW (Figure 14b) and at 1005 m, there is fluctuation of winds from southwestern to northern sector (Figure 14c). This vertical change of wind direction between 105 m and 525 m, is also observed at the second examined case (14 April), where the PBL was very shallow and did not exceed 600 m. The domination of W winds at 105 m and the vertical shift at the first 500 m above ground level, suggests strong layering in the lower levels. More specifically, in Figure 15, the 2-h mean PBL height is calculated for each product: there is no strong daytime evolution and the mean PBL does not exceed 400 m, demonstrating the low layers’ dominance in Finokalia station. During 04:00–08:00 and 16:00–18:00 UTC, MLH_TKE present larger variability comparing to the other 2-h averages. PBL_RCS and PBL_BSC, present the highest deviation at 12:00–14:00 UTC, in addition to the smaller deviations during the rest of the 24 h. The nighttime measured PBL is overestimated by PBL_ECMWFSEA and underestimated by PBL_ECMWFLAND, while both overestimate the observed daytime PBL.

Mean Bias Error (MBE), standard deviation (σ) and coefficient of determination (R²) are calculated for the differences between PBL retrievals from different products for 1–26 April 2017 (Figure 16). The absolute value of MBE ranges from 0.2 m to 43.9 m, but only two out of nine comparisons exceed 20 m, which given the resolution of PollyXT (7.5 m) and Halo (30 m) Lidar, can be characterized as reliable MBE. Least-squares regression is used to derive the linear fit between the 15min PBL height estimates. As observed, there is reasonably good correlation PBL_WVMR-PBL_RCS, PBL_BSC-PBL_RCS and PBL_BSC-PBL_WVMR (R² = 0.92, R² = 0.92, R² = 0.9, respectively) for nighttime measurements and PBL_BSC-PBL_RCS (R² = 0.93) for daytime measurements during 1–26 April. Cases where TKE_dr indicates MLH_TKE < 120 m, were replaced with MLH_TKE = 60 m, corresponding to half the “lower detection limit” of the instrument, in order to avoid positive or negative biases by excluding or zeroing, respectively. Hence, strong deviations occurred in the comparisons that included PBL retrieved from TKE_dr (MLH_TKE—PBL_BSC, MLH_TKE—PBL_RCS, MLH_TKE—PBL_WVMR), with standard deviation exceeding σ = 223. This is expected, given the different nature of direct observations of turbulence and tracers of turbulent mixing as well as previous observations [47]. The systematic difference seen in this inter-comparison, is not attributed to the different methods used but with the different products: WCT is used for the detection of PBL height, using aerosols (RCS, BSC, WVMR) as tracers, while TM is used for the detection of PBL height, using a thermodynamic parameter (TKE_dr) as tracer.

In addition to this, the differences of PBL_WVMR-PBL_RCS (σ = 94.7), PBL_BSC-PBL_RCS (σ = 85.7), PBL_BSC-PBL_WVMR (σ = 101) during nighttime and from PBL_BSC-PBL_RCS (σ = 87.4) during daytime, are described by smaller standard deviation.

4. Conclusions

Finokalia station is placed on a coastal steep cliff, presenting local meteorological characteristics that decisively influence the PBL formation and evolution. Statistical analysis shows that W winds along the coast are dominant in the lower levels most of the time during PreTECT Campaign, forming a shallow PBL that does not present sharp daytime evolution. In addition, when N winds meet the coastline, the forming PBL is clearly higher.

In this study, the cases examined in detail, represent the two different meteorological conditions as described above. It is remarkable that during a sunny day (14 April 2017) the PBL is shallower in comparison to a cloudy one (10 April 2017). The reason for this difference, is that PBL above Finokalia, appears to be dominated by coastal flows rather than thermal convection. More specifically, the strong winds along the coast on a sunny day (14 April), are capping the development of the PBL.

Comparing our observations with the ECMWF results from a sea surface bin and a land surface bin, lead to the conclusion that in both cases, there is an overestimation of the PBL height relative to the estimates derived from lidar, with the exception of nighttime hours (00:00–06:00, 18:00–24:00 UTC), where the land bin underestimates the observed PBL. A coastal region with complex topography such as Finokalia, where convection plays a much smaller role in PBL evolution, favors the formation of a PBL that is not well described by the relatively coarse scale model. The model-measurements differences are addressed to the influence of changes in surface roughness in combination with the horizontal advection of air across this discontinuity. This also emphasizes the importance of the actual PBL measurements when the models are uncertain in such regions.

Overall, the WCT method applied on RCS, WVMR, and BSC, is a trustworthy reference for the detection of layering. The synergistic use of these products is crucial for the determination of PBL height. WVMR is a considerable tracer for the boundary layer dynamics, since it is strongly related to wind transport. Especially in coastal areas, the effect of water surfaces on the evolution of PBL is very well recorded when considering water vapor profiles. RCS and BSC are proportional to the aerosol concentration and constitute a reliable indicator for layering detection. However, different tracers for PBL, may lead to distinct results, but contribute to building a more accurate image of the boundary layer diurnal evolution. Applying TM on TKE_dr profiles, captures the well-mixed layers with thermal and mechanical turbulence, whereas WCT captures a less turbulent layer that contains aerosols (case 10 April 2017 18:00–24:00 UTC). Thus, care must be taken in interpreting the tracer-based PBL height, as the captured layer can differ from the nocturnal layer or the MLH.

In some cases, the mesoscale circulation that causes the strong winds in the Aegean Sea, superimposes on the mean wind flow in Finokalia, and probably resembles a sea-breeze-type circulation. At the end of the day, wind speed and direction change and probably relate to land-breeze-type circulation. These conditions favor the development of a TIBL, stable during nighttime, mainly represented by MLH_TKE.

A possible future research study, could be the creation of an automated algorithm that retrieves the PBL height in Finokalia, given an established pattern of the winds interfering with the PBL height behavior and the development of an IBL.